Quản lý tài nguyên hướng tiết kiệm năng lượng trên môi trường điện toán đám mây

114 9 0
Quản lý tài nguyên hướng tiết kiệm năng lượng trên môi trường điện toán đám mây

Đang tải... (xem toàn văn)

Tài liệu hạn chế xem trước, để xem đầy đủ mời bạn chọn Tải xuống

Thông tin tài liệu

ĐẠI HỌC QUỐC GIA TP HỒ CHÍ MINH TRƯỜNG ĐẠI HỌC BÁCH KHOA NGUYỄN QUANG HÙNG QUẢN LÝ TÀI NGUYÊN HƯỚNG TIẾT KIỆM NĂNG LƯỢNG TRÊN MƠI TRƯỜNG ĐIỆN TỐN ĐÁM MÂY LUẬN ÁN TIẾN SĨ KỸ THUẬT TP HỒ CHÍ MINH NĂM 2019 ĐẠI HỌC QUỐC GIA TP HCM TRƯỜNG ĐẠI HỌC BÁCH KHOA NGUYỄN QUANG HÙNG QUẢN LÝ TÀI NGUYÊN HƯỚNG TIẾT KIỆM NĂNG LƯỢNG TRÊN MÔI TRƯỜNG ĐIỆN TOÁN ĐÁM MÂY Chuyên ngành: Mã số chuyên ngành: KHOA HỌC MÁY TÍNH 62.48.01.01 Phản biện độc lập 1: Phản biện độc lập 2: PGS TS Nguyễn Tấn Khôi TS Nguyễn Đình Minh Phản biện 1: Phản biện 2: Phản biện 3: PGS TS Trần Công Hùng PGS TS Lê Trung Quân PGS TS Huỳnh Xuân Hiệp NGƯỜI HƯỚNG DẪN KHOA HỌC PGS TS Thoại Nam TS Nguyễn Thanh Sơn LỜI CAM ĐOAN Tác giả xin cam đoan cơng trình nghiên cứu thân tác giả Các kết nghiên cứu kết luận luận án trung thực, không chép từ nguồn hình thức Việc tham khảo nguồn tài liệu (nếu có) thực trích dẫn ghi nguồn tài liệu tham khảo quy định Tác giả luận án Chữ ký i TÓM TẮT LUẬN ÁN Điện tốn đám mây (ĐTĐM), mơ hình tính tốn phân bố cỡ lớn, trở thành mơ hình điện tốn tiện ích (utility computing) định hướng tính kinh tế Các đám mây hạ-tầng-như-một-dịch-vụ (IaaS Cloud) cung cấp tài nguyên cho người dùng (đám mây) dạng máy ảo ngày phổ biến trung tâm liệu (TTDL) đám mây Một nghiên cứu ước tính cơng suất tiêu thụ chi phí điện tiêu thụ cho trung tâm liệu Mỹ khoảng 50 MW 15 triệu đô-la năm, nghiên cứu khác chi phí phải trả cho điện tiêu thụ trung tâm liệu ảo hóa ngày tăng xu chi phí lượng tiêu thụ tiếp tục tăng chi phí cho phần cứng khơng đổi Bài toán lập lịch máy ảo hướng tiết kiệm lượng hướng giải cần thiết trung tâm liệu ảo hóa đám mây nhà cung cấp dịch vụ ĐTĐM để giảm chi phí hoạt động Một thách thức toán giảm tổng điện tiêu thụ máy vật lý toán lập lịch máy ảo tiết kiệm lượng tiêu thụ máy vật lý TTDL ảo hóa đám mây (tiếng Anh: Energy-aware scheduling/placement of virtual machines in Cloud virtualized data centers) Luận án nghiên cứu toán lập lịch máy ảo hướng hiệu lượng đám mây hạtầng-như-một-dịch-vụ đám mây tính tốn hiệu cao, máy vật lý có tài nguyên giới hạn (mỗi máy vật lý đặt tối đa số giới hạn g ≥ máy ảo đồng thời) máy ảo yêu cầu nhiều loại tài nguyên khoảng thời gian (bắt đầu, kết thúc), không di dời không nhường thời gian thực thi máy ảo Bài toán lập lịch máy ảo hướng hiệu lượng NP-hard chủ đề nghiên cứu quan tâm Các đóng góp luận án gồm: Bài toán lập lịch máy ảo hướng tiết kiệm lượng tiêu thụ nêu luận án có tính chiều thời gian sử dụng máy ảo máy ảo yêu cầu nhiều loại tài nguyên đồng thời Luận án khai thác khoảng thời gian thực thi máy ảo (thời gian bắt đầu thời gian kết thúc máy ảo) để từ tìm mục tiêu lập lịch tương đương là: “Tối thiểu tổng thời gian bận rộn máy vật lý”; đồng thời sử dụng liệu tải sử dụng chiều tài nguyên máy vật lý yêu ii cầu máy ảo Thay tối thiểu số máy vật lý dùng, luận án đề xuất hai nhóm giải thuật: i Sử dụng chiều thời gian giải pháp lập lịch hướng tiết kiệm lượng luận án đề xuất nhóm giải thuật lập lịch thứ MinDFT (bao gồm giải thuật MinDFT-ST, MinDFT-FT, MinDFT-LDTF MinDFT-LFT chương 5) Các giải thuật MinDFT-ST, MinDFT-FT, MinDFT-LDTF, MinDFT-LFT khác cách xếp danh sách máy ảo (đầu vào) trước gán máy ảo lên máy vật lý giải thuật MinDFT Giải thuật MinDFT heuristic dạng Best-Fit Decreasing (BFD) gán máy ảo lên máy vật lý có thời gian hồn thành máy vật lý tăng thêm nhỏ để tối thiểu tổng lượng tiêu thụ máy vật lý Giải thuật MinDFT có độ phức tạp thời gian O(n.m) với n số máy ảo cần lập lịch m số máy vật lý ii Giải pháp lập lịch hướng tiết kiệm lượng dựa yếu tố thời gian độ dài vector tài nguyên dư (sau cấp cho máy ảo) máy vật lý, luận án đề xuất nhóm giải thuật thứ hai EMinTRE với giải thuật gồm: EMinTRE-ST, EMinTRE-FT, EMinTRE-LDTF EMinTRELFT (ở chương 6) Giải thuật EMinTRE đề xuất sử dụng độ đo (ký hiệu: TRE – different Time and Resource Efficiency) để gán máy ảo Độ đo TRE bao gồm tỉ lệ độ chênh lệch thời gian bận rộn máy vật lý (có nhân với trọng số thời gian) độ dài vector tài nguyên dư máy vật lý (mỗi tài nguyên nhân tỉ lệ phần trăm tài nguyên dư với trọng số) Máy vật lý có giá trị TRE nhỏ chọn để gán máy ảo Giải thuật EMinTRE heuristic dạng Best-Fit Decreasing nhằm tối thiểu tổng thời gian bận rộn máy vật lý dẫn đến rút giảm tổng lượng tiêu thụ máy vật lý Các giải thuật EMinTRE-ST, EMinTRE-FT, EMinTRE-LDTF EMinTRE-LFT khác cách xếp danh sách máy ảo (đầu vào) trước gán máy ảo lên máy vật lý giải thuật EMinTRE Giải thuật EMinTRE có độ phức tạp thời gian O(n.m) với n số máy ảo cần lập lịch m số máy vật lý (giả sử m ≥ logn) Luận án thực đánh giá hiệu giải thuật lập lịch đề xuất mô dựa phần mềm CloudSim (Phần mềm CloudSim phần mềm mô giải thuật cấp phát tài nguyên mơi trường điện tốn iii đám mây phổ biến nhất) Nhiều đánh giá dựa mô thực sử dụng mơ hình tải song song (parallel workload models) Parallel Workload Archive (PWA) cho thấy rằng: “Các giải thuật lập lịch đề xuất (trong luận án) giảm tổng điện tiêu thụ so với giải thuật lập lịch phổ biến khác như: giải thuật Power-Aware Best-Fit Decreasing (ký hiệu: PABFD) A Beloglazov, giải thuật Modified First-fit Decreasing Earliest (ký hiệu: Tian-MFFDE) W Tian, heuristic đóng thùng dạng Norm-based Greedy (ký hiệu: VBP-Norm-L1, VBP-Norm-L2),…” iv ABSTRACT Cloud computing, which is a large-scale distributed computing paradigm, has been becoming an utility computing model and is driven by economies of scale Infrastructure-as-a-Service (IaaS) cloud provides cloud users with computing resources in terms of virtual machines (VMs) to run their applications in cloud virtualized data centers A study has estimated the energy cost of a single data center is more than $15M per year The power consumption is increased with the increasing scale of these data centers Therefore, advanced scheduling techniques for reducing energy consumption of these cloud systems are highly concerned for any cloud providers to reduce energy cost This thesis investigates the energy-efficient virtual machine scheduling problems in IaaS clouds where users request multiple resources in fixed intervals and nonpreemption for processing their virtual machines and physical machines have bounded capacity resources Many previous works are based on migration techniques to move on-line VMs from low utilization hosts and turn these hosts off to reduce energy consumption However, the techniques for migration of VMs could not use in our research The scheduling problem is NP-hard and is one of the hot topic research The key contributions are: Thesis proposes new approach to the problem of energy-aware virtual machine scheduling with the execution time of virtual machines and each virtual machine requires multiple computing resources simultaneously Thesis exploits the execution time of each virtual machine (between the start time and end time of each virtual machine) to find out the equivalent scheduling goal is "Minimizing total busy time of the physical machines" Instead of minimizing the number used physical machines, the thesis propose scheduling algorithms to minimize the sum of total busy time of all physical machines that is equivalent to minimize total energy consumption The thesis proposes two approaches: i Using the execution time of each virtual machine the thesis proposes the Minimizing Differential Finishing Time (MinDFT) algorithms, which is denoted as MinDFT-ST, MinDFT-FT, MinDFT-LDTF and MinDFT-LFT (in the chapter 5) All of MinDFT-ST, MinDFT-FT, MinDFT-LDTF and MinDFT-LFT algorithms differ in sorting of list of virtual machines to be v allocated by core MinDFT algorithm The MinDFT algorithm is a BestFit Decreasing heuristic to allocate a virtual machine to a physical machine that has enough available resources and minimum of the increasingly finishing time of the physical machine The time complexity of MinDFT algorithm is O(n.m) where n is the number of virtual machines to be scheduled, m is the number of physical machines ii The energy-aware virtual machine scheduling algorithms based on both the time and length (Euclidean norm) of multiple dimensional vector where each dimension of the vector is the residual resource utilization (after allocation for placed virtual machines) of a physical machine The thesis proposed the Energy-aware Minimizing Differential Time and Resource Efficiency algorithms, which are denoted as EMinTRE-ST, EMinTRE-FT, EMinTRE-LDTF and EMinTRE-LFT (in Chapter 6) The EMinTRE-ST, EMinTRE-FT, EMinTRE-LDTF and EMinTRE-LFT algorithms are proposed using a new metric (denoted as TRE – different Time and Resource Efficiency) to allocate virtual machines to physical machines The TRE metric is sum of squared of ratio of increasing finishing time of a physical machine (multiplied by the time weight) and length of the residual resources vector of the physical machine (each resource multiplied by percentage of residual resources with weight) A physical machine that has enough available resources and minimum of the TRE value will be selected to allocate a virtual machine Core EMinTRE algorithm is Best-Fit Decreasing on the TRE The EMinTRE-ST, EMinTRE-FT, EMinTRE-LDTF and EMinTRE-LFT algorithms differ in sorting of list of virtual machines to be allocated by the core EMinTRE algorithm The EMinTRE algorithm has the time complexity is O(n.m) where n is the number of virtual machines to be scheduled, m is the number of physical machines (assume that m ≥ log n) Thesis evaluates performance of these proposed scheduling algorithms by simulations Thesis chooses CloudSim, which is a popular Cloud data centers and virtual machine allocation algorithms simulation software There are many extensive simulations using parallel workload models in Parallel Workload vi Archive show that the proposed algorithm has the least total energy consumption compared to the state-of-the-art algorithms include Beloglazov’s Power-Aware Best-Fit Decreasing (denoted as PABFD), Tian’s Modified First-fit Decreasing Earliest (denoted as Tian-MFFDE), vector bin-packing heuristics (denoted as VBP-Norm-L1, VBP-Norm-L2) vii LỜI CÁM ƠN Tôi xin trân trọng cám ơn PGS TS Thoại Nam TS Nguyễn Thanh Sơn tận tình hướng dẫn thực nghiên cứu Tôi xin cảm ơn đồng nghiệp môn Hệ Thống & Mạng Máy Tính, khoa Khoa Học & Kỹ Thuật Máy Tính, phịng Đào tạo Sau đại học trường Đại học Bách Khoa – Đại học Quốc Gia Tp Hồ Chí Minh, thầy nhà khoa học phản biện đóng góp ý kiến cho Luận án Tiến sĩ Tôi xin cảm ơn đến GS TS Josef Küng trường Johannes Kepler University of Linz/JKU, Austria giúp đỡ nhiều thời gian nghiên cứu trường học bổng Erasmus Mundus GATE project Con chân thành gửi đến cha: Nguyễn Văn Nghiệm mẹ: Nguyễn Thị Thu Đông biết ơn sinh thành, dưỡng dục Tôi gửi lời cảm ơn đến vợ: Hồ Kim Oanh hai Nguyễn Kim Anh Nguyễn Ngọc Vân Anh đồng hành thời gian qua viii OPT 𝐸 (𝒱, 𝑔) ≥ a 𝑃0)1 𝑙𝑒𝑛(𝒱 ) 𝑔 Chứng minh cận tổng điện tối thiểu sau Áp dụng cận biên chi phí tối ưu từ quan sát 7.1, có: OPT(𝒱, g) ≤ len(𝒱) Do ta có cận tổng điện tối thiểu: 𝐸 (𝒱, 𝑔) ≤ a 𝑃0)1 𝑙𝑒𝑛(𝒱) + Z 𝑒, OPT (7.5) ,@' Giả sử công suất tiêu thụ máy vật lý tuyến tính theo tải CPU sử dụng Áp dụng phương trình tính lượng tiêu thụ chương 4, ta có lượng tiêu thụ máy ảo thứ i suốt khoảng thời gian [si, si + duri], giả sử công suất tiêu thụ tăng lên từ tải sử dụng CPU máy vật lý tăng từ Ui lên 𝑈,S sau gán máy ảo thứ i lên là: I-$7$ 𝑒, = ∫2 $ $ I-$7$ 𝑃! (𝑈, → 𝑈,S )𝑑𝑡 = ∫2 $ $ (1 − 𝛼 ) 𝑃0)1 𝑢, 𝑑𝑡=(1 − 𝛼) 𝑃0)1 𝑢, 𝑑𝑢𝑟, Trong ui tải sử dụng CPU máy ảo thứ i Bởi máy ảo bị ràng buộc điều kiện tối đa 100% tải CPU máy vật lý hay ≤ ui ≤ Do đó: 𝑒, ≤ (1 − 𝛼) 𝑃0)1 𝑑𝑢𝑟, Gọi Ii khoảng thời gian máy ảo thứ i, Ii= [si, si + duri] Theo định nghĩa (4.1) độ dài khoảng Ii len(Ii) = (si + duri) – si = duri Do đó: 𝑒, ≤ (𝑃0)1 − 𝑃,-./ ) 𝑙𝑒𝑛(𝐼, ) Tổng lượng tiêu thụ n máy ảo là: ∑4,@' 𝑒, ≤ ∑4,@'[(𝑃0)1 − 𝑃,-./ ) 𝑙𝑒𝑛(𝐼, )] Theo định nghĩa, tổng chiều dài khoảng n máy ảo là: 𝑙𝑒𝑛(𝒱 ) = ∑4,@' 𝑙𝑒𝑛(𝐼, ) Nên: Z 𝑒, ≤ (𝑃0)1 − 𝑃,-./ ) 𝑙𝑒𝑛(𝒱) ,@' Thế (7.6) vào (7.5) ta có: OPT 𝐸 (𝒱, 𝑔) ≤ 𝑃,-./ 𝑙𝑒𝑛(𝒱 ) + (𝑃0)1 − 𝑃,-./ ) 𝑙𝑒𝑛(𝒱 ) Vậy: 81 (7.6) OPT 𝐸 (𝒱, 𝑔) ≤ 𝑃0)1 𝑙𝑒𝑛(𝒱) (7.7) Kết hợp (7.5) (7.7) ta chứng minh định lý (7.3) Định lý 7.2: Cho tập máy ảo 𝒱 (tương ứng với tập 𝓙 các khoảng thời gian) máy vật lý có tối đa g máy ảo gán đồng thời (trong thời điểm), giải thuật gồm: MinDFT-LDTF, MinDFT-LFT, EMinTRE-LDTF EMinTRE-LFT đề xuất có độ xấp xỉ g Nghĩa tổng thời gian bận rộn lập lịch cho tập 𝒱 tạo giải thuật MinDFT-LDTF, MinDFT-LFT, EMinTRELDTF EMinTRE-LFT nhỏ hay g lần tổng thời gian bận rộn lập lịch tối ưu tập 𝒱 Ta có: EMinTRE-LFT(𝒱 ) ≤ 𝑔.OPT(𝒱, 𝑔) (7.8) EMinTRE-LDTF(𝒱 ) ≤ 𝑔.OPT(𝒱, 𝑔) (7.9) MinDFT-LDTF(𝒱) ≤ 𝑔.OPT(𝒱, 𝑔) (7.10) MinDFT-LFT(𝒱) ≤ 𝑔.OPT(𝒱, 𝑔) (7.11) Chứng minh: Đặt n(t) số máy ảo lập lịch thời điểm t, mA(t) số máy vật lý cần dùng giải thuật EMinTRE-LFT thời điểm t, mopt(t) số máy vật lý cần dùng thời điểm t lập lịch tối ưu Với thời điểm t ≥ toàn lập lịch giải thuật EMinTRE-LFT, máy vật lý gán máy ảo, nên: ≤ mA(t) ≤ n(t) n(t) ≥ Rỏ ràng thời điểm t ≥ số lượng máy ảo tối đa gán lên máy vật lý lớn hay 4(%) (g ≥ 0) nên 𝑚6#% (𝑡) ≥ 𝑚6#% (t) ≥ 0& (%) M M ≥ 0& (%) M Do đó: ⟺ 𝑚 + (𝑡) ≤ 𝑔 𝑚6#% (𝑡) g ≥ độc lập với thời gian t Lấy tích phân hai vế ta (nhắc lại: 𝑚 + (𝑡) ≥ 0, 𝑚6#% (t) ≥ 0): 2#)4(𝒱) 2#)4(𝒱) + G 𝑚 (𝑡)𝑑𝑡 ≤ 𝑔 G N N 82 𝑚6#% (𝑡)𝑑𝑡 (7.12) Giải thuật EMinTRE-LFT lập lịch cho tập 𝒱 có tổng thời gian bận rộn tồn lập lịch ký hiệu EMinTRE-LFT(𝒱) Theo định nghĩa 7.2 EMinTRE-LFT(𝒱) tích phân hàm mA(t) khoảng thời gian t ∈ [0, span(𝒱)] EMinTRE-LFT(𝒱) = G 2#)4(𝒱) 𝑚 + (𝑡)𝑑𝑡 (7.13) N Thế (6.17) vào (6.16) ta có: EMinTRE-LFT(𝒱 ) ≤ 𝑔 G 2#)4(𝒱) 𝑚6#% (𝑡)𝑑𝑡 (7.14) N Vậy: EMinTRE-LFT(𝒱 ) ≤ 𝑔.OPT(𝒱, 𝑔) Chứng minh tương tự cho giải thuật MinDFT-ST, MinDFT-FT, MinDFT-LDTF MinDFT-LFT với bất đẳng thức: MinDFT-FT(𝒱 ) ≤ 𝑔.OPT(𝒱, 𝑔), MinDFT-ST(𝒱 ) ≤ 𝑔.OPT(𝒱, 𝑔), MinDFT-LDTF(𝒱 ) ≤ 𝑔.OPT(𝒱, 𝑔), MinDFT-LFT(𝒱 ) ≤ 𝑔.OPT(𝒱, 𝑔), EMinTRE-ST(𝒱 ) ≤ 𝑔.OPT(𝒱, 𝑔), EMinTRE-FT(𝒱 ) ≤ 𝑔.OPT(𝒱, 𝑔), EMinTRE-LDTF(𝒱 ) ≤ 𝑔.OPT(𝒱, 𝑔) Định lý chứng minh Kết luận Chương đánh giá độ hiệu giải thuật đề xuất mặt lý thuyết Trong trường hợp tổng quát giải thuật đề xuất có tổng thời gian bận rộn lập lịch cho tập 𝒱 (các máy ảo) tạo hai giải thuật MinDFT EMinTRE nhỏ hay g lần tổng thời gian bận rộn lập lịch tối ưu tập 𝒱 83 CHƯƠNG TỔNG KẾT Tóm tắt Mục tiêu luận án để giải toán lập lịch máy ảo máy vật lý mục tiêu tiết kiệm lượng tiêu thụ máy vật lý Thách thức toán thuộc lớp toán NP-hard với số chiều tài nguyên lớn Trước đề xuất giải thuật lập lịch hướng tiết kiệm lượng cho trung tâm liệu đám mây, nghiên cứu sinh phân tích tốn lập lịch máy ảo với máy ảo yêu cầu nhiều loại tài nguyên thực thi khoảng cố định, máy vật lý có tài ngun giới hạn Bài tốn khác tốn đóng thùng nhiều chiều (ddimensional vector bin-packing problem) Mặc dù giải thuật First-fit Decreasing hay Best-Fit Decreasing tốn đóng thùng (bin-packing problem) tối thiểu số máy vật lý dùng, luận án phương pháp tối thiểu số máy vật lý không đồng nghĩa với việc tối thiểu tổng điện tiêu thụ máy vật lý Giả sử máy vật lý hoạt động khoảng thời gian bắt đầu kết thúc cố định điện tiêu thụ máy vật lý tuyến tính theo tải sử dụng CPU việc tối thiểu tổng thời gian bận rộn máy vật lý dẫn đến tối thiểu tổng điện tiêu thụ máy vật lý Khai thác đặc điểm máy ảo yêu cầu thực thi khoảng thời gian thực thi cố định, không di dời, không nhường máy ảo cần nhiều loại tài nguyên để thực thi luận án đề xuất giải thuật lập lịch tốt giải thuật lập lịch có Sử dụng chiều thời gian giải pháp lập lịch hướng tiết kiệm lượng luận án đề xuất nhóm giải thuật lập lịch thứ MinDFT (bao gồm giải thuật MinDFTST, MinDFT-FT, MinDFT-LDTF MinDFT-LFT chương 5) Nhóm giải thuật lập lịch thứ MinDFT (bao gồm giải thuật MinDFT-ST, MinDFT-FT, MinDFTLDTF MinDFT-LFT chương 5) tối thiểu tổng thời gian thực thi tăng thêm máy vật lý Tất giải thuật nhóm thứ dựa giải thuật MinDFT gán máy ảo lên máy vật lý có tổng thời gian hồn thành tăng thêm nhỏ Các giải thuật MinDFT-ST, MinDFT-FT, MinDFT-LDTF MinDFT-LFT khác cách xếp danh sách máy ảo trước gán MinDFT-ST danh sách máy ảo theo 84 thời gian bắt đầu sớm trước, MinDFT-FT danh sách máy ảo theo thời gian kết thúc sớm trước, MinDFT-LDTF danh sách máy ảo theo thời gian thực thi dài trước MinDFT-LFT danh sách máy ảo theo thời gian kết thúc sau trước Khi dựa vào thời gian tăng thêm nhỏ để gán máy ảo đến máy vật lý dẫn đến việc sử dụng tài nguyên máy vật lý chưa tối ưu như: gán máy ảo dùng nhiều nhớ lên máy vật lý dẫn đến khơng cịn nhớ để gán tiếp cho máy ảo khác tương lai, lựa chọn khác gán lên máy vật lý khác có tổng thời gian tăng thêm tối thiểu tài nguyên sử dụng tốt Đó lý nhóm giải thuật thứ hai EMinTRE với giải thuật gồm: EMinTRE-ST, EMinTRE-FT, EMinTRE-LDTF EMinTRE-LFT (ở chương 6) đề xuất Giải pháp lập lịch hướng tiết kiệm lượng dựa yếu tố thời gian độ dài vector tài nguyên dư (sau cấp cho máy ảo) máy vật lý, luận án đề xuất nhóm giải thuật thứ hai EMinTRE với giải thuật gồm: EMinTRE-ST, EMinTRE-FT, EMinTRE-LDTF EMinTRE-LFT (ở chương 6) Tất giải thuật nhóm thứ hai EMinTRE dựa giải thuật EMinTRE để tối thiểu tổng thời gian bận rộn máy vật lý quan tâm đến việc tài nguyên hiệu gán máy ảo lên máy vật lý Luận án đề xuất độ đo TRE (Time increasing – Resource Efficency) dựa độ dài vector tài nguyên trống thời gian bận rộn tăng thêm máy vật lý Ý nghĩa độ đo TRE gán máy ảo lên máy vật lý quan tâm thêm chiều tài nguyên khác (như CPU, nhớ, băng thông mạng,…) dùng Nếu máy có thời gian bận rộn tăng thêm giải thuật gán máy ảo lên máy vật lý có độ dài vector tài ngun trống cịn lại nhỏ Tất giải thuật EMinTRE-ST, EMinTRE-FT, EMinTRE-LDTF EMinTRE-LFT gán máy ảo lên máy vật lý có giá trị TRE nhỏ nhất, giải thuật khác cách xếp danh sách máy ảo trước gán đến máy vật lý giải thuật EMinTRE Giải thuật EMinTRE-ST xếp danh sách máy ảo theo thời gian bắt đầu tăng dần, giải thuật EMinTRE-FT xếp danh sách máy ảo theo thời gian kết thúc tăng dần, giải thuật EMinTRE-LDTF xếp danh sách máy ảo theo thời gian thực thi yêu cầu máy ảo dài trước, giải thuật EMinTRE-LDTF xếp danh sách máy ảo theo thời gian kết thúc sau trước 85 Để đánh giá độ hiệu giải thuật đề xuất này, luận án trình bày phương pháp đánh giá giải thuật thực nghiệm mô giải thuật đề xuất giải thuật so sánh phần mềm CloudSim, phần mềm mơ mơi trường điện tốn đám mây giải thuật phân bổ máy ảo môi trường điện toán đám mây nhiều nghiên cứu sử dụng rộng rãi Thông tin tải công việc (workload) cho hệ thống sử dụng tải công việc thực Parallel Workload Archive từ mơ hình tải công việc song song (parallel workload models [77]) nhiều nghiên cứu sử dụng rộng rãi nhóm tác giả: Feitelson cộng [37], Lublin cộng [36], Downey và cộng [38],… Kết mô cho thấy so sánh với giải thuật lập lịch khác có, hai nhóm giải thuật MinDFT EMinTRE luận án đề xuất cho kết tốt (tổng điện tiêu thụ hơn) giải thuật lập lịch khác có Power-Aware Best-Fit Decreasing (PABFD) A Beloglazov cộng [25][13], heuristic đóng thùng VBP-Norm-L2 [27], Modified First-Fit Decreasing Earliest (Tian-MFFDE) W Tian cộng [23] Các giải thuật EMinTRE cho phép lựa chọn trọng số tài nguyên thời gian công thức TRE Việc chọn lựa trọng số cho chiều tài ngun thời gian cơng thức tính TRE máy vật lý gán máy ảo chọn cách xếp danh sách máy ảo thích hợp với workload cho kết tốt (tổng lượng tiêu thụ nhỏ hơn) giải thuật EMinTRE Hiện mô gán trọng số tài nguyên thời gian Trong nghiên cứu mở rộng tương lai kết hợp với kỹ thuật học máy (machine learning) để chọn trọng số tài nguyên phù hợp với loại hệ thống máy vật lý dạng workload khác Luận án chứng minh cận độ xấp xỉ lý thuyết giải thuật đề xuất MinDFT-ST, MinDFT-FT, MinDFT-LDTF, MinDFT-LFT, EMinTRE-ST, EMinTREFT, EMinTRE-LDTF EMinTRE-LFT so với giải thuật tối ưu không số g (với g số máy ảo tối đa đặt máy vật lý thời điểm bất kỳ) Các kết trình bày chương Trong vài trường hợp đặc biệt hiệu giải thuật lập lịch đề xuất tốt (nếu khoảng máy vật lý khơng chứa – proper interval graph giải thuật EMinTRE-LDTF có độ xấp xỉ & (3 − M) so với tối ưu) 86 Hướng nghiên cứu mở rộng Từ nghiên cứu kết đạt luận án này, nghiên cứu sinh đề nghị số vấn đề hướng nghiên cứu sau: - Vấn đề 1: Trong công thức tính độ đo TRE cần biết trọng số tài nguyên Luận án giả định tài nguyên có trọng số (và 1) thực đánh giá mô Trong nghiên cứu gần [78], nghiên cứu sinh tập thể hướng dẫn áp dụng phương pháp mạng trí tuệ nhân tạo giải thuật di truyền tìm trọng số tương quan với danh sách máy ảo yêu cầu (đầu vào) Tuy nhiên tải cơng việc đầu vào thay đổi việc học lại giá trị trọng số tài nguyên chưa khảo sát cơng trình [78] - Vấn đề 2: Các máy vật lý bất đồng môi trường điện toán đám mây Luận án khảo sát máy vật lý đồng Trường hợp máy vật lý bất đồng nghiên cứu sinh tập thể hướng dẫn đề xuất EPOBF [34] dựa tổng số MIPS Watt để chọn máy vật lý gán máy ảo, giải thuật EM [20] kết hợp gán máy ảo theo thời gian bận rộn nhỏ độ đo tổng số MIPS Watt để chọn máy vật lý gán máy ảo, giải thuật di truyền GAPA [21] để gán máy ảo lên máy vật lý Tuy nhiên vấn đề với máy vật lý bất đồng cần khảo sát kỹ hơn, quan tâm nhiều khả tài ngun băng thơng mạng ảnh hưởng hiệu của ứng dụng song song (có sử dụng giao tiếp qua mạng) - Vấn đề 3: Các ứng dụng tính tốn hiệu cao có giao tiếp qua mạng Luận án dừng lại việc mô tả máy ảo cần băng thông mạng chưa cho thấy quan tâm đến topology tiến trình ứng dụng song song, chưa khảo sát hạ tầng mạng máy vật lý dùng cơng nghệ mạng tốc độ cao đặc biệt (như Infiniband) - Vấn đề 4: Mơ hình lượng tiêu thụ máy vật lý Luận án quan tâm đến lượng tiêu thụ máy vật lý Năng lượng tiêu thụ trung tâm liệu gắn với lượng hệ thống làm lạnh, thiết bị mạng, tủ rack,… 87 DANH MỤC CÔNG TRÌNH ĐÃ CƠNG BỐ N Quang-hung, N Thoai, and N T Son, “Energy-saving Virtual Machine Scheduling in Cloud Computing with Fixed Interval Constraints,” Transactions Large-Scale Data- Knowledge-Centered Systems XXXI, SpringerVerlag, Selected Papers from FDSE 2015, pp 124-145, ISSN: 1869-1994, 2017 N Quang-Hung and N Thoai, “Minimizing Total Busy Time with Application to Energy-efficient Scheduling of Virtual Machines in IaaS Clouds,” in Proceedings of the International Conference on Advanced Computing and Applications (ACOMP 2016), IEEE, 2016, pp 141-150 T Nguyen-Duc, N Quang-Hung, N Thoai, “BFD-NN: Best Fit DecreasingNeural Network for Online Energy-aware Virtual Machine Allocation Problems”, in Proceedings of the Seventh Symposium on Information and Communication Technology (SoICT 2016), ACM, 2016, pp 243-250 N Quang-Hung and N Thoai, “EMinRET: Heuristic for Energy-Aware VM Placement with Fixed Intervals and Non-preemption,” in Proceedings of the International Conference on Advanced Computing and Applications (ACOMP 2015), IEEE, 2015, pp 98-105 N Quang-Hung and N Thoai, “Minimizing Total Busy Time for Energy-Aware Virtual Machine Allocation Problems,” in Proceedings of the Sixth International Symposium on Information and Communication Technology (SoICT 2015), ACM, 2015, pp 179-186 N Quang-Hung, D Le, N Thoai, and N T Son, “Heuristics for Energy-Aware VM Allocation in HPC Clouds,” in Proceedings of the International Conference on Future Data and Security Engineering (FDSE 2014), Springer-Verlag, 2014, ISBN 978-3-319-12777-4, DOI: 10.1007/978-3-319-12778-1_19, pp 248-261 N Quang-Hung, N Thoai, and N T Son, “EPOBF: Energy Efficient Allocation of Virtual Machines in High Performance Computing Cloud,” Transactions Large-Scale Data- Knowledge-Centered Systems XVI, Springer-Verlag, Selected Papers from ACOMP 2013, Vol 8960, pp 71-86, Oct 2014 N Quang-Hung, N Thoai, N T Son, and D Le, “Energy-Aware Lease Scheduling in Virtualized Data Centers”, The 5th International Conference on High Performance Scientific Computing (HPSC 2012), March 5-9, 2012, Hanoi, Vietnam, Selected Papers and Invited Talks of HPSC 2012, Springer-Verlag, 2014, ISBN: 978-3-319-09062-7, pp 195-205 N Quang Hung, N Thoai, and N T Son, “Performance constraint and poweraware allocation for user requests in virtual computing lab,” Journal of Science and Technology (Vietnam), vol 49, No 4A, Special on International Conference on Advanced Computing and Applications (ACOMP 2011), 2011, pp 383-392 88 TÀI LIỆU THAM KHẢO [1] M Armbrust, A D Joseph, R H Katz, and D A Patterson, “Above the Clouds : A Berkeley View of Cloud Computing,” in Technical Report No UCB/EECS-2009-28, 2009, pp 1–23 [2] R Buyya, R Buyya, C S Yeo, C S Yeo, S Venugopal, S Venugopal, J Broberg, J Broberg, I Brandic, and I Brandic, “Cloud computing and emerging IT platforms: Vision, hype, and reality for delivering computing as the 5th utility,” Futur Gener Comput Syst., vol 25, no 6, pp 599–616, Jun 2009 [3] R Buyya, C S Yeo, and S Venugopal, “Market-Oriented Cloud Computing: Vision, Hype, and Reality for Delivering IT Services as Computing Utilities,” 2008 10th IEEE Int Conf High Perform Comput Commun., pp 5–13, Sep 2008 [4] B Sotomayor, R S Montero, I M Llorente, and I Foster, “Virtual Infrastructure Management in Private and Hybrid Clouds,” IEEE Internet Comput., vol 13, no 5, pp 14–22, Sep 2009 [5] T C Ferreto, M A S Netto, R N Calheiros, and C A F De Rose, “Server consolidation with migration control for virtualized data centers,” Futur Gener Comput Syst., vol 27, no 8, pp 1027–1034, 2011 [6] S K Garg, C S Yeo, A Anandasivam, R Buyya, C S Yeo, A Anandasivam, and R Buyya, “Environment-conscious scheduling of HPC applications on distributed Cloud-oriented data centers,” J Parallel Distrib Comput., vol 71, no 6, pp 732–749, Jun 2010 [7] Í Goiri, J L Berral, J O Fitó, F Julià, R Nou, J Guitart, R Gavaldà, and J Torres, “Energy-efficient and multifaceted resource management for profitdriven virtualized data centers,” Futur Gener Comput Syst., vol 28, no 5, pp 718–731, May 2012 [8] I I Goiri, F Julia, R R Nou, J L Berral, J Guitart, J Torres, Jordi Torres, J Torres, Í Goiri, F Julià, R R Nou, J L Berral, and F Juli, “Energy-Aware Scheduling in Virtualized Datacenters,” in 2010 IEEE International Conference on Cluster Computing, 2010, pp 58–67 [9] K Le, R Bianchini, J Zhang, Y Jaluria, J Meng, and T D Nguyen, “Reducing electricity cost through virtual machine placement in high performance computing clouds,” in Proceedings of 2011 International Conference for High Performance Computing, Networking, Storage and Analysis on - SC ’11, 2011, p 22:1 22:12 [10] M A Salehi, P R Krishna, K S Deepak, and R Buyya, “Preemption-Aware Energy Management in Virtualized Data Centers,” in 2012 IEEE Fifth International Conference on Cloud Computing, 2012, pp 844–851 [11] B Sotomayor, “Provisioning Computational Resources Using Virtual Machines and Leases,” The University of Chicago, Chicago, US, 2010 [12] X Fan, W D Weber, and L A Barroso, “Power provisioning for a warehouse89 sized computer,” in Proceedings of the 34th annual international symposium on Computer architecture, 2007, vol 2, no June, pp 13–23 [13] A Beloglazov, J H Abawajy, and R Buyya, “Energy-aware resource allocation heuristics for efficient management of data centers for Cloud computing,” Futur Gener Comput Syst., vol 28, no 5, pp 755–768, May 2012 [14] A Beloglazov, R Buyya, Y C Lee, and A Zomaya, “A Taxonomy and Survey of Energy-Efficient Data Centers and Cloud Computing Systems,” Adv Comput., vol 82, pp 1–51, 2011 [15] Y Ding, X Qin, L Liu, and T Wang, “Energy efficient scheduling of virtual machines in cloud with deadline constraint,” Futur Gener Comput Syst., vol 50, pp 62–74, 2015 [16] T Knauth and C Fetzer, “Energy-aware scheduling for infrastructure clouds,” in 4th IEEE International Conference on Cloud Computing Technology and Science Proceedings, 2012, pp 58–65 [17] A Gupta, L V Kale, D Milojicic, P Faraboschi, and S M Balle, “HPC-Aware VM Placement in Infrastructure Clouds,” in 2013 IEEE International Conference on Cloud Engineering (IC2E), 2013, pp 11–20 [18] I Takouna, W Dawoud, and C Meinel, “Energy efficient scheduling of HPCjobs on virtualized clusters using host and VM dynamic configuration,” ACM SIGOPS Oper Syst Rev., vol 46, no 2, p 19, Jul 2012 [19] G B Mertzios, M Shalom, A Voloshin, P W H Wong, and S Zaks, “Optimizing busy time on parallel machines,” Theor Comput Sci., vol 562, no May 2012, pp 524–541, 2015 [20] N Quang-Hung and N Thoai, “Minimizing Total Busy Time for Energy-Aware Virtual Machine Allocation Problems,” in The Sixth International Symposium on Information and Communication Technology - SoICT 2015, 2015, pp 179–186 [21] N Quang-Hung, P D P Nien, N N H Nam, N H Tuong, and N Thoai, “A Genetic Algorithm for Power-Aware Virtual Machine Allocation in Private Cloud,” in ICT-EurAsia’13, vol 7804, K Mustofa, E J Neuhold, A M Tjoa, E Weippl, and I You, Eds Berlin, Heidelberg: Springer-Verlag Berlin Heidelberg, 2013, pp 183–191 [22] N Quang-Hung and N Thoai, “EMinRET: Heuristic for Energy-Aware VM Placement with Fixed Intervals and Non-preemption,” in 2015 International Conference on Advanced Computing and Applications (ACOMP), 2015, pp 98– 105 [23] W Tian and C S Yeo, “Minimizing total busy time in offline parallel scheduling with application to energy efficiency in cloud computing,” Concurr Comput Pract Exp., vol 27, no 9, pp 2470–2488, Jun 2013 [24] A Beloglazov, “Energy-Efficient Management of Virtual Machines in Data Centers for Cloud Computing,” Department of Computing and Information Systems, University of Melbourne, 2013 [25] A Beloglazov and R Buyya, “Optimal online deterministic algorithms and 90 adaptive heuristics for energy and performance efficient dynamic consolidation of virtual machines in Cloud data centers,” Concurr Comput Pract Exp., vol 24, no 13, pp 1397–1420, Sep 2012 [26] M R Garey and D S Johnson, Computers and Intractability: A Guide to the Theory of NP-Completeness New York, NY, USA: W H Freeman & Co., 1974 [27] R Panigrahy, K Talwar, L Uyeda, and U Wieder, “Heuristics for vector bin packing,” research.microsoft.com, 2011 [28] C Chekuri and S Khanna, “On Multi-dimensional Packing Problems,” in SODA ’99 Proceedings of the tenth annual ACM-SIAM symposium on Discrete algorithms, 1999, pp 185–194 [29] L Epstein and A Levin, “Bin packing with general cost structures,” Math Program., vol 132, no 1–2, pp 355–391, 2012 [30] A.-C Orgerie, M D de Assuncao, and L Lefevre, “A survey on techniques for improving the energy efficiency of large-scale distributed systems,” ACM Comput Surv., vol 46, no 4, p 47:1 47:31, Mar 2014 [31] S Srikantaiah, A Kansal, and F Zhao, “Energy aware consolidation for cloud computing,” in Proceedings of the 2008 conference on Power aware computing and systems, 2008, pp 10–10 [32] L Chen and H Shen, “Consolidating complementary VMs with spatial/temporal-awareness in cloud datacenters,” in Proceedings - IEEE INFOCOM, 2014, pp 1033–1041 [33] R Buyya, A Beloglazov, and J Abawajy, “Energy-efficient management of data center resources for cloud computing: A vision, architectural elements, and open challenges,” arXiv Prepr arXiv1006.0308, pp 1–12, 2010 [34] N Quang-Hung, N Thoai, and N T Son, “EPOBF: Energy Efficient Allocation of Virtual Machines in High Performance Computing Cloud,” Trans LargeScale Data- Knowledge-Centered Syst XVI Sel Pap from ACOMP 2013, vol 8960, pp 71–86, Oct 2014 [35] M Flammini, G Monaco, L Moscardelli, H Shachnai, M Shalom, T Tamir, and S Zaks, “Minimizing total busy time in parallel scheduling with application to optical networks,” Theor Comput Sci., vol 411, no 40–42, pp 3553–3562, Sep 2010 [36] U Lublin and D G Feitelson, “The workload on parallel supercomputers: Modelling the characteristics of rigid jobs,” J Parallel Distrib Comput., vol 63, no 11, pp 1105–1122, Nov 2003 [37] D G Feitelson, “Packing schemes for gang scheduling,” in Job Scheduling Strategies for Parallel Processing, 1996, pp 89–110 [38] A B Downey, “A parallel workload model and its implications for processor allocation,” in Proceedings The Sixth IEEE International Symposium on High Performance Distributed Computing (Cat No.97TB100183), 1997, pp 112–123 91 [39] D G Feitelson, “‘Parallel workloads archive’ URL http://www.cs.huji.ac.il/labs/parallel/workload/.” [40] R N Calheiros, R Ranjan, A Beloglazov, C A F De Rose, and R Buyya, “CloudSim: a toolkit for modeling and simulation of cloud computing environments and evaluation of resource provisioning algorithms,” Softw Pract Exp., vol 41, no 1, pp 23–50, Jan 2011 [41] “The San Diego Supercomputer Center (SDSC) Blue Horizon log-trace (SDSCBLUE-2000-4.1-cln.swf.gz file).” [42] D Feitelson, “HPC2N Seth log (HPC2N-2002-2.2-cln.swf),” 2014 [Online] Available: http://www.cs.huji.ac.il/labs/parallel/workload/l_hpc2n/HPC2N2002-2.2-cln.swf.gz [43] P Mell and T Grance, “The NIST Definition of Cloud Computing,” NIST, vol 800–145, pp 1–7, Jan 2011 [44] R Buyya, J Broberg, and A Goscinski, Cloud Computing: Principles and Paradigms Hoboken, NJ, USA: John Wiley & Sons, Inc., 2011 [45] T V Pham, H Jamjoom, K E Jordan, and Z.-Y Shae, “A service composition framework for market-oriented high performance computing cloud,” in Proceedings of the 19th ACM International Symposium on High Performance Distributed Computing - HPDC ’10, 2010, p 284 [46] B Sotomayor, S Montero, and I Mart, “Capacity Leasing in Cloud Systems using the OpenNebula Engine,” in Workshop on Cloud Computing and its Applications 2008 (CCA08), 2008 [47] J Y.-T Leung, Handbook of Scheduling: Algorithms, Models, and Performance Analysis, no Chapman and Hall/CRC, 2004 [48] S Dasgupta, C H Papadimitriou, and U Vazirani, Algorithms McGraw-Hill, Inc., 2006 [49] D G Feitelson, L Rudolph, and U Schwiegelshohn, “Parallel Job Scheduling — A Status Report,” in Computing, 2005, pp 1–16 [50] Y Liu and H Zhu, “A survey of the research on power management techniques for high-performance systems,” Softw Pract Exp., vol 40, no 11, pp 943–964, 2010 [51] G L Valentini, W Lassonde, S U Khan, N Min-Allah, S A Madani, J Li, L Zhang, L Wang, N Ghani, J Kolodziej, H Li, A Y Zomaya, C.-Z Xu, P Balaji, A Vishnu, F Pinel, J E Pecero, D Kliazovich, and P Bouvry, “An overview of energy efficiency techniques in cluster computing systems,” Cluster Comput., vol 16, no 1, pp 3–15, Sep 2011 [52] A Beloglazov and R Buyya, “Adaptive threshold-based approach for energyefficient consolidation of virtual machines in cloud data centers,” in Proceedings of the 8th International Workshop on Middleware for Grids, Clouds and eScience - MGC ’10, 2010, pp 1–6 [53] R Ge, X Feng, and K W Cameron, “Performance-constrained Distributed 92 DVS Scheduling for Scientific Applications on Power-aware Clusters,” in SC ’05 Proceedings of the 2005 ACM/IEEE conference on Supercomputing, 2005, p 34 [54] K H Kim, A Beloglazov, and R Buyya, “Power-aware provisioning of virtual machines for real-time Cloud services,” Concurr Comput Pract Exp., vol 23, no 13, pp 1491–1505, 2011 [55] S Albers and H Fujiwara, “Energy-efficient algorithms for flow time minimization,” ACM Trans Algorithms, vol 3, no 4, p 49:1-49:17, Nov 2007 [56] S Albers, “Energy-efficient algorithms,” Commun ACM, vol 53, no 5, p 86, May 2010 [57] G von Laszewski, L Wang, A J Younge, and X He, “Power-aware scheduling of virtual machines in DVFS-enabled clusters,” in 2009 IEEE International Conference on Cluster Computing and Workshops, 2009, pp 1–10 [58] B Speitkamp and M Bichler, “A Mathematical Programming Approach for Server Consolidation Problems in Virtualized Data Centers,” IEEE Trans Serv Comput., vol 3, no 4, pp 266–278, Oct 2010 [59] M H Ferdaus, M Murshed, R N Calheiros, and R Buyya, “Virtual Machine Consolidation in Cloud Data Centers Using ACO Metaheuristic,” in Euro-Par 2014 Parallel Processing, 2014, vol 8632, pp 306–317 [60] D Borgetto, M Maurer, G Da-Costa, J.-M Pierson, and I Brandic, “Energyefficient and SLA-aware management of IaaS clouds,” in Proceedings of the 3rd International Conference on Future Energy Systems Where Energy, Computing and Communication Meet - e-Energy ’12, 2012, pp 1–10 [61] A Beloglazov and R Buyya, “Energy Efficient Resource Management in Virtualized Cloud Data Centers,” in 2010 10th IEEE/ACM International Conference on Cluster, Cloud and Grid Computing, 2010, pp 826–831 [62] R N Calheiros, R Ranjan, C A F De Rose, and R Buyya, “CloudSim : A Novel Framework for Modeling and Simulation of Cloud Computing Infrastructures and Services,” Architecture, pp 1–9 [63] O Mämmelä, M Majanen, R Basmadjian, H Meer, A Giesler, and W Homberg, “Energy-aware job scheduler for high-performance computing,” Comput Sci - Res Dev., vol 27, no 4, pp 265–275, Aug 2011 [64] Y Kessaci, N Melab, and E.-G Talbi, “A Pareto-based metaheuristic for scheduling HPC applications on a geographically distributed cloud federation,” Cluster Comput., vol 16, no 3, pp 451–468, May 2012 [65] L A Barroso, J Clidaras, and U Hölzle, “The datacenter as a computer: An introduction to the design of warehouse-scale machines,” Synth Lect Comput Archit., vol 8, no 3, pp 1–154, 2013 [66] L A Barroso and U Hölzle, “The Case for Energy-Proportional Computing,” Computer (Long Beach Calif)., vol 40, no 12, pp 33–37, Dec 2007 [67] G Chen, W He, J Liu, S Nath, L Rigas, L Xiao, and F Zhao, “Energy-Aware 93 Server Provisioning and Load Dispatching for Connection-Intensive Internet Services,” in USENIX Symposium on Networked Systems Design and Implementation (NSDI), 2008, pp 337–350 [68] N Q Hung, N Thoai, and N T Son, “Performance constraint and power-aware allocation for user requests in virtual computing lab,” J Sci Technol., vol 49, 4A, no Special on International Conference on Advanced Computing and Applications (ACOMP2011), pp 409–418, 2011 [69] C Ghribi, M Hadji, and D Zeghlache, “Energy Efficient VM Scheduling for Cloud Data Centers: Exact Allocation and Migration Algorithms,” in 2013 13th IEEE/ACM International Symposium on Cluster, Cloud, and Grid Computing, 2013, pp 671–678 [70] N Quang-Hung, L T Tan, C T Phat, and N Thoai, “A GPU-Based Enhanced Genetic Algorithm for Power-Aware Task Scheduling Problem in HPC Cloud,” in ICT-EurAsia 2014, Second IFI., vol LNCS 8407, Linawati, M S Mahendra, E J Neuhold, A M Tjoa, and I You, Eds Berlin, Heidelberg: Springer Berlin Heidelberg, 2014, pp 159–169 [71] B Sotomayor, K Keahey, and I Foster, “Combining batch execution and leasing using virtual machines,” in Proceedings of the 17th international symposium on High performance distributed computing - HPDC ’08, 2008, pp 87–96 [72] N Quang-Hung, D.-K Le, N Thoai, and N T Son, “Heuristics for EnergyAware VM Allocation in HPC Clouds,” in Future Data and Security Engineering: First International Conference, FDSE 2014, Ho Chi Minh City, Vietnam, November 19-21, 2014, 2014, vol LNCS 8860, pp 248–261 [73] Z Á Mann, “Rigorous results on the effectiveness of some heuristics for the consolidation of virtual machines in a cloud data center,” 2015 [74] D Feitelson, “Parallel Workloads Archive http://www.cs.huji.ac.il/labs/parallel/workload/.” [75] M Dabbagh, B Hamdaoui, M Guizani, and A Rayes, “Release-time aware VM placement,” in 2014 IEEE Globecom Workshops (GC Wkshps), 2014, pp 122– 126 [76] M Y Kovalyov, C T Ng, and T C E Cheng, “Fixed interval scheduling: Models, applications, computational complexity and algorithms,” Eur J Oper Res., vol 178, no 2, pp 331–342, Apr 2007 [77] D Feitelson, “Workload modeling for performance evaluation,” Perform Eval Complex Syst Tech Tools, pp 114–141, 2002 [78] T Nguyen-Duc, N Quang-Hung, and N Thoai, “BFD-NN: Best Fit DecreasingNeural Network for Online Energy-Aware Virtual Machine Allocation Problems,” in Proceedings of the Seventh Symposium on Information and Communication Technology - SoICT ’16, 2016, pp 243–250 [79] B Varghese and R Buyya, “Next generation cloud computing: New trends and research directions,” Futur Gener Comput Syst., vol 79, no September, pp 94 849–861, Feb 2018 [80] A Hameed et al., “A survey and taxonomy on energy efficient resource allocation techniques for cloud computing systems,” Computing, vol 98, no 7, pp 751–774, Jul 2016 [81] Y Sharma, B Javadi, W Si, and D Sun, “Reliability and energy efficiency in cloud computing systems: Survey and taxonomy,” J Netw Comput Appl., vol 74, pp 66–85, Oct 2016 [82] C Mobius, W Dargie, and A Schill, “Power Consumption Estimation Models for Processors, Virtual Machines, and Servers,” IEEE Trans Parallel Distrib Syst., vol 25, no 6, pp 1600–1614, Jun 2014 [83] M Zakarya and L Gillam, “Energy efficient computing, clusters, grids and clouds: A taxonomy and survey,” Sustain Comput Informatics Syst., vol 14, pp 13–33, Jun 2017 [84] B Jennings and R Stadler, “Resource Management in Clouds: Survey and Research Challenges,” J Netw Syst Manag., vol 23, no 3, pp 567–619, Jul 2015 [85] A Vogel, D Griebler, C A F Maron, C Schepke, and L G Fernandes, “Private IaaS Clouds: A Comparative Analysis of OpenNebula, CloudStack and OpenStack,” in 2016 24th Euromicro International Conference on Parallel, Distributed, and Network-Based Processing (PDP), 2016, pp 672–679 95 ... thùng Tiết kiệm lượng Tiết kiệm lượng Tiết kiệm lượng Đánh đổi (trade-off) Tiết kiệm lượng tỉ lệ chấp nhận công việc Tiết kiệm lượng Y Kessaci cộng [64] Có CPU Tian Yeo [23] Có Đơn tài nguyên. .. CHƯƠNG CƠ SỞ LÝ THUYẾT Điện toán đám mây (Cloud computing) 2.1.1 Điện toán đám mây (cloud computing) Điện toán đám mây, viết tắt ĐTĐM, Rajkumar Buyya cộng [2] định nghĩa: “Một đám mây kiểu hệ...ĐẠI HỌC QUỐC GIA TP HCM TRƯỜNG ĐẠI HỌC BÁCH KHOA NGUYỄN QUANG HÙNG QUẢN LÝ TÀI NGUYÊN HƯỚNG TIẾT KIỆM NĂNG LƯỢNG TRÊN MƠI TRƯỜNG ĐIỆN TỐN ĐÁM MÂY Chuyên ngành: Mã số chuyên ngành:

Ngày đăng: 28/02/2021, 21:41

Từ khóa liên quan

Mục lục

  • TrangBia-LATS

  • NQHUNG-LATS_HDCaptruong-ver1_3b

Tài liệu cùng người dùng

  • Đang cập nhật ...

Tài liệu liên quan